Features/Start activity from another activity
- 1 Summary
- 2 Owner
- 3 Current status
- 4 Detailed Description
- 5 Benefit to Sugar
- 6 Scope
- 7 UI Design
- 8 How To Test
- 9 User Experience
- 10 Dependencies
- 11 Contingency Plan
- 12 Documentation
- 13 Release Notes
- 14 Comments and Discussion
This feature provides a new way of interoperability between activities, the current ones being the Journal and the Clipboard.
Activities are provided with a new API to call other activities. This API allows the expression of phrases such as "read this book" or "edit this code with that editor".
This is important for activities that handle sets of objects. For example a gallery of images, a content catalog.
- Name: Manuel Quiñones
- Targeted release: 0.106
- Last updated: May, 2015
- Percentage of completion: 100%
One of Sugar key principles is simplicity. Activities are meant to be simple, and users can make interesting things happen when they interoperate with them. Currently, they do so by using the Journal and the Clipboard.
However, there are cases where you want to start an activity from inside another, thus skipping the step through the Journal.
This feature provides a third way of interoperability between activities.
For example, an activity whose purpose is to download content, would like to display the content right away.
The following considerations are made.
- This solution is based on current activities, ie does not imply a change in activities. It uses the available information that Sugar has now:
- the bundle registry
- the activity declared mime types
- the metadata in datastore objects
- This solution is for every bundle, not just activities. This includes content bundles.
- This solution considers Rainbow security.
An activity can start other activity by:
- knowing the activity ID - starts that specific activity
- passing a datastore object - if the activity ID is also passed, starts the object with that specific activity. Otherwise starts the first activity that can handle the object.
- passing a mime type - if the activity ID is not passed, start the first activity that can handle the mime type.
An error is reported to the user of the API if the request couldn't be accomplished, with a message. For example "no activity can start this object" or "activity with ID XYZ not found".
The code that starts activities in Sugar currently lives in the Journal. The idea is to move the code to a more general module in Sugar, and make the Journal a user of it. Then, add a DBUS service so that the activitie's toolkit can use the API too.
Benefit to Sugar
This will enhace the interoperability of Sugar activities, removing the need to pass through the Journal in the cases where it makes sense and feels indirect or complicated. Thus, is an enhacement of the user experience.
The lack of this capability made Uruguay deployment (Ceibal) go for a custom solution regarding content obtention and consumption. They mixed the capabilities of activities GetBooks, Read and Jukebox into a monolithic activity "Biblioteca Segura".
This feature affects only Sugar modules. Current activities can take advantage of the new API provided by this feature.
The Journal is adapted to be a user of this API. In fact, all the code is adapted from the Journal code.
Video that shows the user experience: https://dl.dropboxusercontent.com/u/588753/start.webm
There is no UI in this feature. Activities that currently have a button "Show in Journal" can move to this API and change the button label to "Start" or "Display" or something more specific like "Read Book" or "Edit Code".
How To Test
- Ensure Image Viewer activity is installed,
- Ensure Browse-157.1 is installed, if not use browse-157.1.xo,
- Start Browse and navigate to the Wiki,
- Move the cursor over an image, such as Media:Sugarlabs_mainpage_01.png, right-click and select Save Image As, wait for the download to complete,
- Verify that the alert offers Open with Image Viewer,
- Click on Open with Image Viewer,
- Verify that the Image Viewer activity is started and the image is shown.
UX enhacement is described in the section above "Benefit to Sugar".
This feature does not depend on the completion of other features.
In 2011 Martin Abente (tch) sent a patch that enhaced sugar-launch http://marc.info/?l=olpc-sugar&m=129485435106175&w=2
As discussed in the mailing list, this solution is not appropiate because it skips Rainbox security (http://wiki.laptop.org/go/Rainbow) in secured systems, and seems odd that the child activity process is spawned from the parent activity process. Sugar "the shell" should spawn the activity processes.
Android Intents as reference
I loosely based this feature in Android Intents:
Android has a similar mechanism than the proposed here. Is much more powerful and used system-wide and also internally per application. Android apps are made of components, and those components declare which actions can perform as intent filters. Using a data structure called Intent, one component can call an action of another component. That data stracture holds an abstract description of the action to be performed. Android has an intent resolution mechanism to determine the component that matches the intent.
The main similarities and differences are:
- Currently we can only have one action: "Start" that launches an activiy. Our activities are not made of components like Android apps. If that were the case, we could be able to do neat things like "take a photo" from any activity, that would call the correspondent component of Record or other available activity to take a photo from the camera. This is a major change that implies a change in activities, so is out of the scope of this feature.
- Android Intents use a URI for data, usually a content provider like "content://..". The current way to reference an object in Sugar is by its ID. Currently, we can't pass arbitrary data to an activity.
- In Android, the component to be called can be implied from the data. We do the same. Our components are the Sugar bundles.
- In Android, extra data can be passed. We can't do that in current activities.
- In Android, the mime type can be explicit or implied from the data. We do the same.
To be written.
Comments and Discussion
Your comments are very welcome. Please discuss through the mailing list.
- First discussion on the mailing list: